Diagnostics of Mechanical Faults in Power Transformers - Vibration Sensor Network Design under Vibration Uncertainty

Abstract

Power transformer is a critical component in energy transmission, and its failure can cause catastrophic social loss. Among many techniques to prevent the transformer failures, ones using vibration signals show good capability of detecting the mechanical faults. For on-site power transformers, numerous vibration sensors are installed to take into account vibration uncertainty which comes from sizable and complex transformers and random operating condition. It, however, brings about the high maintenance cost of sensing system as well as superfluous data obstructing precise diagnostics. This study proposes sensor positioning to detect mechanical faults of power transformers. Thirty six on-site power transformers in nuclear power plants were employed. Their vibration signals are processed based upon the principles of transformer vibration. Vibration characteristics are analyzed in terms of spectrum analysis, vibration contour plot and high vibration locations. Then the sensor network design framework is proposed which adjusts the number of sensors and their locations to measure high vibration signals robustly under vibration uncertainty. It is demonstrated that the designed sensing system evaluates the health status of the power transformers successfully with the significantly reduced number of sensors.